Comments on: The Five Percent Rule: So Far So Right http://seedround.com/the-five-percent-rule-so-far-so-right The Adventures of Boxxet, Startup Life, and the Web. Sun, 23 Nov 2008 15:13:26 +0000 http://wordpress.org/?v=2.6.2 By: David http://seedround.com/the-five-percent-rule-so-far-so-right#comment-1112 David Thu, 16 Nov 2006 02:50:54 +0000 http://seedround.com/the-five-percent-rule-so-far-so-right#comment-1112 Perhaps the 5% is the level of creativity given a blank page. What if systems (Boxxet included) were designed to prompt creativity along chosen lines or through multiple methods. The 5% may be a result of the structure, not the content of the individuals. Change the structure and get 15% and you have outdone your assumptions by 300%. Just a thought. Perhaps the 5% is the level of creativity given a blank page. What if systems (Boxxet included) were designed to prompt creativity along chosen lines or through multiple methods. The 5% may be a result of the structure, not the content of the individuals. Change the structure and get 15% and you have outdone your assumptions by 300%. Just a thought.

]]> By: Niraj Sanghvi http://seedround.com/the-five-percent-rule-so-far-so-right#comment-39 Niraj Sanghvi Fri, 17 Mar 2006 19:33:34 +0000 http://seedround.com/the-five-percent-rule-so-far-so-right#comment-39 Seems like the filtering and preference steps are very closely linked (as performing opposite functions). I imagine the filtering mechanism must be pretty complex then. I look forward to seeing it in action as the participation/userbase grows. Seems like the filtering and preference steps are very closely linked (as performing opposite functions). I imagine the filtering mechanism must be pretty complex then. I look forward to seeing it in action as the participation/userbase grows.

]]> By: You Mon Tsang http://seedround.com/the-five-percent-rule-so-far-so-right#comment-38 You Mon Tsang Fri, 17 Mar 2006 18:26:32 +0000 http://seedround.com/the-five-percent-rule-so-far-so-right#comment-38 Sure, Niraj: The participation within Boxxet does a few things. 1) SORTING: It manually sorts the current information (higher-rated goes up the list) 2) FILTERING: The system learns how to filter the content better, so marking a irrelevant story on, say, the TV show "24" can teach the system not to let that type of story in later. 3) PREFERENCE: Also participants may prefer certain kinds of content, certain sources, or certain story types. Boxxet will learn from that and rank accordingly the new stuff it finds. I am glad you are finding it easy to participate. That is our goal. As our user base settles down, we would love to drive the participation rate up. But like you said, 5% is realistic and we should plan for no more than that. Sure, Niraj:

The participation within Boxxet does a few things.

1) SORTING: It manually sorts the current information (higher-rated goes up the list)
2) FILTERING: The system learns how to filter the content better, so marking a irrelevant story on, say, the TV show “24″ can teach the system not to let that type of story in later.
3) PREFERENCE: Also participants may prefer certain kinds of content, certain sources, or certain story types. Boxxet will learn from that and rank accordingly the new stuff it finds.

I am glad you are finding it easy to participate. That is our goal. As our user base settles down, we would love to drive the participation rate up. But like you said, 5% is realistic and we should plan for no more than that.

]]> By: Niraj Sanghvi http://seedround.com/the-five-percent-rule-so-far-so-right#comment-37 Niraj Sanghvi Fri, 17 Mar 2006 18:17:54 +0000 http://seedround.com/the-five-percent-rule-so-far-so-right#comment-37 I absolutely agree about the 5 percent creating. I'm just starting to use boxxet and finding it is fairly easy to participate. Could you elaborate any on what you mean when you say the system learns based on participation? Does it use participation to improve its own filters? I absolutely agree about the 5 percent creating. I’m just starting to use boxxet and finding it is fairly easy to participate. Could you elaborate any on what you mean when you say the system learns based on participation? Does it use participation to improve its own filters?

]]>